Atomic layer deposition of alumina on porous polytetrafluoroethylene membranes for enhanced hydrophilicity and separation performances

Journal of Membrane Science 415–416 (2012) 435–443

Contents lists available at SciVerse ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Atomic layer deposition of alumina on porous polytetrafluoroethylene membranes for enhanced hydrophilicity and separation performances Qiang Xu a, Yang Yang b, Xiaozu Wang a, Zhaohui Wang a, Wanqin Jin a, Jun Huang a, Yong Wang a,n a b

State Key Laboratory of Materials-Oriented Chemical Engineering, College of Chemistry and Chemical Engineering, Nanjing University of Technology, Nanjing 210009, PR China Laboratory for Nanotechnology, Institute of Microsystems Engineering (IMTEK), Albert-Ludwigs-University Freiburg, Freiburg 79110, Germany

a r t i c l e i n f o

abstract

Article history: Received 19 March 2012 Received in revised form 10 May 2012 Accepted 15 May 2012 Available online 26 May 2012

Porous polytetrafluoroethylene (PTFE) membranes are seeking extensive applications in gas and liquid purification whereas the strong hydrophobicity of PTFE limits their use in water system. We demonstrate that atomic layer deposition (ALD) is a simple and effective method to upgrade the separation performances of PTFE membranes by improving their surface hydrophilicity. Alumina was ALD-deposited on PTFE membranes following a subsurface nucleation and surface growth mechanism, leading to the formation of fine particulates with lower ALD cycles and continuous thick layers with higher ALD cycles on the surface of PTFE membranes, respectively. Thanking to the nanofibrilcontaining, porous surface morphology, the adhesion between the deposited alumina and PTFE substrate is strong, which can withstand harsh ultrasonication. The hydrophilicity of the deposited membranes was enhanced progressively with the rise of cycle numbers and the membrane with 500 cycles was completely water-wettable with a water angle less than 201. Improved hydrophilicity of the alumina-deposited PTFE membranes not only affords remarkably enhanced fouling resistance, but also facilitates water permeation through the membrane, giving higher flux. Filtration experiments indicate that the deposited membranes prepared at optimal conditions possessed an increase of pure water flux of more than 50% and simultaneously an increase of retention of 12.4% compared to the pristine PTFE membrane. & 2012 Elsevier B.V. All rights reserved.

Keywords: Atomic layer deposition PTFE Membrane Surface modification

1. Introduction Polytetrafluoroethylene (PTFE) is a highly symmetrical nonpolar linear polymer material with a helical conformation, in which a skeleton constructed of carbon atoms is surrounded by fluorine atoms [1,2]. Due to its good thermal stability, superior chemical resistance, potential biocompatibility, high mechanical strength and low dielectric constant [3–6], PTFE is extensively used in diverse fields from our daily life to industries. PTFE porous membranes are manufactured using a stretching process and enjoy great success as filters for air and solvent purification. However, when PTFE membranes come to deal with filtration and separation in water, the strong hydrophobicity of PTFE significantly degrades its performances. First of all, strong hydrophobicity prevents water from penetrating into the PTFE membrane, which not only requires higher pressure consuming more energy, but also reduces flux. Even worse, because of the strong hydrophobic interaction between PTFE with hydrophobic solutes in water, PTFE membranes are easily fouled by absorption, resulting

n

Corresponding author. Tel.: þ86 25 8317 2247; fax: þ 86 25 8317 2292. E-mail address: [email protected] (Y. Wang).

0376-7388/$ - see front matter & 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.memsci.2012.05.031

in decreasing flux, frequent membrane regeneration, and shorter lifetime. There are some efforts to improve the hydrophilicity of PTFE. However, because of its chemical and physical inertness, energetic techniques including electron beam irradiations [7–10], plasma modification [11–13], and chemical treatments [14–16], are typically required to activate the strong C–F bonds, followed by grafting of hydrophilic polymer chains. Apparently, this approach suffers from harsh pretreatment and tedious synthesis and purification protocols. Therefore, it is highly demanded to find a simple route to effectively enhance the hydrophicility of PTFE membranes under moderate operation conditions. Atomic layer deposition (ALD) has been known as an efficient technique for depositing thin films conformally on various substrates with atomic layer control by using sequential surface reactions. Especially, the self-limiting growth mechanism as well as gas phase reaction of ALD enables the deposition of uniform films on the surface of substrates with porous or hierarchical structures. Nucleation and growth of metal oxides such as Al2O3, ZnO, and TiO2 during the ALD process on a variety of polymeric substrates have been investigated in recent years [17,18]. A model based on the adsorption of ALD precursors onto the surface and diffusion into the near-surface region of polymers was proposed for polymers with and without active surface chemical groups,

436

Q. Xu et al. / Journal of Membrane Science 415–416 (2012) 435–443

respectively [17]. Even on PTFE, the most inert polymer, coating of metal oxides by ALD has been achieved recently. It was suggested that the film growth on PTFE initiated from the absorption and retention of the ALD metal precursor within the near surface region of the PTFE films, thus resulting in a more physically rather than a chemically bonded oxide/PTFE interface [18,19]. Kemell et al. investigated the ALD deposition of Al2O3 and TiO2 on nonporous PTFE films [18]. They observed both Al2O3 and TiO2 grew on the PTFE surface as globular particulates. The hydrophilicity of the PTFE films was slightly improved but the adhesion between the deposited layer and PTFE substrates was weak. Very recently, Lee et al. studied the ALD process of ZnO on PTFE tapes, and revealed that although PTFE does not have reactive surface groups, the high reactivity of ALD precursors is possible to lead to an incorporation of transition metals, for example Zn, into PTFE, partly changing the molecular structure of PTFE and affecting its corresponding mechanical properties [19]. Since both the precursor diffusion and the coordination reaction inside the polymer matrix are kinetic-controlled, employment of an exposure process during each ALD half cycle is frequently required for enhanced infiltration efficiency [20]. Inspired by the previous work demonstrating near-surface growth mechanism of metal oxide in PTFE substrate, in this work we investigate ALD of alumina on porous PTFE membranes aiming to improve their surface properties and consequently their separation performances in water system. Moreover, since deposition of oxides onto membrane pores reduces the effective pore size, PTFE membranes subjected to ALD treatment will also have changing retention properties. Porous PTFE membranes produced by the stretching process contain highly oriented fibrils with a diameter ranging from several nanometers to hundreds of nanometers, and the free spaces between these nanofibrils define the pores in the PTFE membrane. Compared to relatively dense PTFE films investigated in previous works, adsorption, diffusion, and reaction of ALD precursors take place predominantly on/in the PTFE nanofibrils, which may show different behaviors and correspondingly changing deposition effects since the morphology, surface area, and orientation of crystallites of PTFE nanofibrils and bulk film differ profoundly. Furthermore, the highly porous nature of PTFE membranes will facilitate the penetration of precursor vapors into the membrane, leading to a fast approaching and adsorption onto PTFE nanofibrils. By taking the advantage of the exposure process, we achieved good control in the deposition of alumina on porous PTFE membranes with adjustable thickness and morphology of the deposition layer. In addition to the effect of continuous pore size tuning, the ALD of alumina effectively improved the hydrophicility of PTFE membranes. As a consequence, the fouling resistance and water flux of the deposited membranes were greatly enhanced.

2. Experimental 2.1. Materials Porous PTFE membranes (Sartorius, Germany) in the form of round chips (diameter: 25 mm; thickness: 65 mm) with a mean pore size of 0.2 mm were used as received for the ALD of alumina. Nonporous PTFE films (Nanjing F4 Chemical Products Co., Ltd) with a thickness of 200 mm were also used as ALD substrates to compare the adhesion of alumina on porous and nonporous PTFE substrates. Monodispered polystyrene nanospheres with a diameter of 190 nm were synthesized by emulsion polymerization [21]. The alumina ALD precursors were trimethylaluminum (TMA, 99.99%) (Organometallics Center, Nanjing University) and deionized H2O, which were used as the metal precursor and the

oxidant source, respectively. N2 with ultrahigh purity (99.99%) was used as both a precursor carrier and the purging gas. Bovine serum albumin (BSA, Mw ¼66 kDa, GM Corporation) with a purity 497% was used as the model protein to compare the fouling resistance of different membrane samples. 2.2. ALD on PTFE membranes PTFE substrate membranes were placed in the holder of a commercialized ALD reactor (Savannah S100, Cambridge NanoTech) and dried at operating temperature for  30 min in vacuum (  1 Torr). Both precursors were maintained in the storage cylinders at room temperature. For a typical ALD cycle, TMA and water vapor were sequentially pulsed into the reactor, and exposure mode [22,23], was applied in this work to ensure sufficient precursor adsorption and diffusion. Thus a total ALD cycle included (i) TMA pulse for 0.015 s; (ii) exposure for 6 s; (iii) purge for 30 s; (iv) H2O pulse for 0.015 s; (v) exposure for 6 s, and (vi) purge for 30 s. ALD was usually carried out at 150 1C for different cycles up to 500 times with a steady N2 flow rate of 10 sccm. In the study of the temperature effect, deposition at 80, 200, and 250 1C was also performed and other deposition conditions were kept unchanged. 2.3. Characterizations of microstructure The surface morphology of membrane samples were examined under a Hitachi S4800 field emission scanning microscope (FESEM) operated at 10 kV. Prior to SEM observations, samples were sputtering-coated with a thin layer of gold. Elemental analysis of the PTFE membrane after alumina deposition was performed at 20 kV using an Oxford INCA 350 energy dispersive X-ray microanalysis system (EDS) equipped to the SEM. In order to analyze alumina impregnated into the PTFE matrix, some deposited membranes were microtomed for TEM observations. Alumina-coated PTFE membranes were embedded in a Spurr lowviscosity epoxy resin, which were then cut into slices with a thickness of about 50–80 nm using a Leica Ultracut microtome equipped with a diamond knife. Slices were collected onto copper grids coated with holey carbon films and probed under a Hitachi H7650 transmission electron microscope (TEM) operating at an accelerating voltage of 80 kV. X-ray diffraction (XRD) patterns of samples prepared at different deposition conditions were obtained from a wide-angle diffractometor with CuKa radiation (l ¼0.154 nm) at a generator voltage of 40 kV and a generator current of 40 mA. The scanning speed and the step were 2.41/min and 0.021, respectively. FTIR spectra were recorded in the transmission configuration on a Thermo Nicolet AVATAR 360 FTIR spectrometer (32 scans, 4 cm  1). Thermogravimetric (TG) analysis was performed with a NETZSCH TG209F1 thermal analyzer under nitrogen atmosphere with a heating rate of 10 1C/min from 30 to 700 1C. Water contact angles were measured on a Dropmeter A-100 contact angle goniometer. For each sample, at least three different sites were measured and the mean value of the three measurements was reported. 2.4. Evaluation the adhesion between the deposited alumina layer and the PTFE substrate A piece of membranes subjected to 50 and 500 ALD cycles were separately immersed in excessive hydrochloric acid solution with a concentration of 5 wt% for 24 h to completely dissolve the deposited alumina layer. The solution containing aluminum species was diluted with deionized water for the detection of aluminum concentration by an inductive coupled plasma emission spectrometer (ICP, Optima 7000DV, Perkin Elmer) to obtain

Q. Xu et al. / Journal of Membrane Science 415–416 (2012) 435–443

the mass of the deposited alumina on one piece of membranes. Another piece of membranes subjected to 50 and 500 ALD cycles was separatedly ultrasonicated under deionized water for 10 min at a power of 300 W and an ultrasound frequency of 40 kHz. Alumina detached from the PTFE substrate was dissolved with excessive hydrochloric acid and aluminum concentration was measured again to know the mass of alumina detached from one piece of membrane. By comparing these two mass values of PTFE membranes subjected to 50 and 500 ALD cycles, we can obtain the percentage of detached alumina which is an indication of the strength of adhesion between the deposited alumina and PTFE substrate. 2.5. Filtration performances Pure water filtration was performed on pristine and deposited membranes with a stirred cell module (Amicon 8010, Millipore) under moderate agitation at room temperature. The hydrophobic membranes including the pristine PTFE membrane and slightly deposited membranes subjected to less than 100 cycles were conditioned with ethanol for 1 min prior to water permeation, while membranes subjected to higher ALD cycles, e.g., 4100 cycles were hydrophilic enough and no pre-wetting was needed. All membranes were first circulated at 2.2 bar for 3.5 h to eliminate the compaction effect to get a constant flux, then the water flux were measured under a pressure of 16.2 kPa. The retention property of different membranes was characterized by filtrating polystyrene nanospheres with a uniform diameter of 190 nm at a pressure of 1.0 bar under moderate agitation at room

437

temperature. We measured the concentration of polystyrene nanospheres in the feed (CF) and in the permeate (CP) by measuring their total amount of carbon using a total organic carbon analyzer (TOC, Japan Shimadzu TOC-VCPH). The retention (R) was given as R¼ 1 CP/CF. For the evaluation of the fouling resistance, pristine and alumina-deposited PTFE membrane were first circulated with pure water at 1.5 bar for 5 h to eliminate the compaction effect and their constant flux in pure water (F0) were measured at 1.0 bar, then we used BSA solution with a concentration of 1.0 g/L prepared in phosphate buffer as the feeding solution for the filtration experiment at 1.0 bar and the flux (F) was recorded every 10 min for 90 min. Values of F/F0 at different sampling points of each sample were plotted with running time to show the trend of flux declining, demonstrating the fouling resistance of the corresponding membranes.

3. Results and discussion 3.1. Surface morphology evolution in alumina ALD on PTFE membranes PTFE membranes manufactured by the stretching process are highly porous, exhibit a net-like structure composed of interconnected nanofibrils (Fig. 1a). The surface of unmodified PTFE nanofibrils is smooth and free of any particulate matter. When subjected to ALD treatment, the PTFE membrane gained small particulates on the surface of the nanofibrils and EDS analysis revealed that Al and O elements appeared on the deposited membrane, indicating that alumina deposition successfully occurred on the PTFE substrate and appeared as surface particulates. Generally, the alumina particulates grew randomly in the form of globular grains on the surface and continuously enlarged with increasing ALD numbers (Fig. 1b–h). For instance, the size of the alumina particulates increased

Fig. 1. SEM images of the surface morphology of the pristine and ALD alumina-coated PTFE membranes: (a) the pristine, (b) 10 cycles, (c) 20 cycles, (d) 50 cycles, (e) 100 cycles, (f) 200 cycles, (g) 300 cycles, and (h) 500 cycles.

438

Q. Xu et al. / Journal of Membrane Science 415–416 (2012) 435–443

from less than 5 nm at 50 cycles to  30 nm at 200 cycles and reached to larger than 100 nm at 500 cycles. Alumina particulates formed below 100 ALD cycles are relatively small and separate from each other, and they started to fuse into continuous and compact layers at higher number of cycles, e.g., 300 and 500 cycles (Fig. 1g and h). For the membrane subjected to 500 ALD cycles, the alumina layer showed a relatively smooth surface morphology (Fig. 1h). The smoothening of the surface subjected to Al2O3 deposition with large cycle numbers is probably due to the conformal feature of ALD films, which can ‘‘repair’’ rough surfaces by filling the space among adjacent particles. Simultaneously, the thickness of the ALD Al2O3 layer gradually increased with increasing ALD numbers, which resulted in a continuing decline of the average pore size of the PTFE membranes. Moreover, some narrow gaps between neighboring nanofibrils disappeared because of the thickening of these nanofibrils with continuous deposition.

3.2. Growth mechanism of alumina ALD on PTFE membranes The formation of the discrete alumina particulates rather than a conformal layer during the initial ALD process on PTFE membranes can be explained by the model proposed by Wilson et al. for polymers with different nature [17]. For polymers with hydroxyl groups on their surface, the precursor molecules such as TMA can react with these reactive surface groups, thus inducing ALD nucleation and further initiating ALD growth. Under this condition, the deposited alumina thin film is easy to be conformal to the polymer substrate. Moreover, the ALD layer chemically bonded to the underlying substrate usually leads to a good adhesion at the film/substrate interface. However, the initial ALD alumina nucleation on polymers lacking reactive surface groups (for instance, PTFE) is different. The deposition proceeds only because TMA molecules can be trapped within the nearsurface region of the PTFE substrate where a larger free volume exists. This assumption is reasonable because PTFE is an extremely hydrophobic nonpolar polymer, which is expected to offer a high chemical solubility to nonpolar TMA [24]. The subsequent pulse of water vapor will react with this absorbed TMA and produce alumina nucleation clusters. It should be mentioned that the regions with a larger free volume under the polymer surface are random and discrete. The retention of TMA in these regions is seriously affected by the period for TMA diffusion and the following purging process. As a result, an initial homogeneous ALD surface nucleation cannot be expected. Some preferential nucleation sites will

favor the formation of alumina clusters therein, which will further evolve to globular particulates and isolated islands during progressive TMA and H2O exposures. After these particulates gradually coalesce and approach to each other, PTFE will be covered by a nearly continuous alumina layer. The normal ALD mode will dominate the subsequent growth of the alumina. In order to reveal the growth mechanism of ALD alumina on porous PTFE membranes, we carried out TEM observations on the ultrathin slices microtomed from PTFE membranes with changing ALD cycles of alumina deposition. As shown in Fig. 2a, there are a few small particles with a diameter o3 nm distributed in slices of the PTFE membrane with 10 alumina ALD cycles. These particles were darker than the surroundings and should be identified as alumina as alumina has a higher electron density than polymeric components. The formation of isolated alumina particulates underneath the surfaces other than continuous thin films along the surface suggests that the ALD growth of alumina on PTFE porous membrane followed a subsurface nucleation mode. For ALD cycles increased to 20, more particles appeared and there were bigger alumina particles with a diameter of  5 nm present adjacent to the surface (Fig. 2b). Furthermore, these particles were dispersed in the PTFE matrix in a gradient way: the number and size of the particles decreased from subsurface to the interior. Therefore, we can conclude that subsurface nucleation remained at least when the cycle number is less than 50, leading to the increase in particle number and size. However, the subsurface nucleation were diffusion-controlled for precursor vapors needed to diffuse into the PTFE matrix to be trapped and reacted, which resulted in a gradient distribution of alumina particles with biggest ones located in the positions nearest to the surface. For the sample subjected to 50 cycles, the gradient distribution of alumina particles is more evident and the size of the particles scattered in different depth under the PTFE surface homogeneously increased compared to their counterparts located in similar positions in the sample with 20 cycles (Fig. 2c). For instance, the diameter of alumina particles located on the surface was  12 nm which was doubled compared to that of the 20-cycle sample. Further increase of cycle numbers to 200 led to a preferential growth of alumina particles adhered to the PTFE surface. Neighboring particles coalesced and merged, forming a thick layer of fused particles with a diameter of  30 nm (Fig. 2d). The fused particulate morphology of the PTFE substrate revealed by TEM agreed well with the SEM observation shown in Fig. 1f. Because of the relatively ‘‘dense’’ layer of alumina, the diffusion of precursors into the PTFE subsurface was restricted and consequently, the subsurface nucleation of alumina were mostly terminated as

Fig. 2. TEM images of microtomed PTFE membranes subjected to changing cycles of Al2O3 ALD at 150 1C: (a) 10 cycles, (b) 20 cycles, (c) 50 cycles, and (d) 200 cycles.

Q. Xu et al. / Journal of Membrane Science 415–416 (2012) 435–443 revealed by the similar number density and particle size of the alumina particles in the subsurface region of the 50- and 200-cycle samples. Therefore, we conclude that in the initial stage of alumina ALD the subsurface nucleation dominated, whereas surface growth on the pre-formed alumina particles played the major role in the late stage. In between the two modes of growth occurred in a balanced way. We measured the mass uptake of the PTFE membranes during the ALD process to compare the deposition rate in different stages with changing growth mode. Thermogravimetric analysis was performed on PTFE membranes subjected to different alumina ALD cycles by degrading the PTFE component in nitrogen up to 700 1C. As shown in Fig. 3a, pristine PTFE started to lose its weight at  550 1C and completely degraded at  600 1C. PTFE membranes subjected to increasing alumina cycles had more remaining weight after the high temperature degradation process because more alumina, which is thermal stable and does not lose weight during the degradation, was deposited onto PTFE membranes. PTFE membranes subjected to 50, 200, and 500 ALD cycles had a remaining weight of 3.12%, 17.4%, and 51.1%, respectively (Fig. 3b–d). Correspondingly, we can calculate the mass uptake per cycle during the period of 0–50 cycle, 50–200 cycle, and 200–500 cycle, which are 0.064%, 0.119%, and 0.278% of the native membrane, respectively. Since the 0–50 cycle and 200–500 cycle cover mostly the subsurface nucleation stage and surface growth stage, respectively, we draw the conclusion that ALD alumina deposition in the surface growth mode at a much higher rate, more than four times larger than that of the subsurface nucleation mode because of the barrier effect of the polymer matrix. When both modes take effect, for instance, in the range of 50–200 cycles in the present case, the deposition rate falls in the window defined by the rate of the subsurface nucleation and the rate of surface growth. Furthermore, the TG curves suggest that alumina deposition does not have a noticeable effect on the thermal stability of the PTFE membranes since both pristine and alumina-deposited PTFE membranes with changing ALD cycles start to be degraded at  550 1C. In addition, prior to the commencement of PTFE degradation, there was a slight weight loss started at  100 1C for the alumina-deposited membranes. This early weight loss should be attributed to the desorption of water molecules bound to the alumina layer deposited on the PTFE substrate. The membrane subjected to 200 alumina cycles had the highest amount of adsorbed water (  8%) because of its rough surface composed of alumina particulates, while the membrane subjected to 50 alumina cycles showed the lowest amount of adsorbed water since the deposited alumina was mostly embedded in the PTFE matrix. Less amount of adsorbed water for the 500-cycle membrane compared to the 200-cycle one should be attributed to the smaller surface area of the former due to the smoother surface composed of fused alumina particulates, which was clearly confirmed in the SEM observations (Fig. 1f and h). 3.3. Influence of the deposited alumina on the crystalline structure of PTFE Fig. 4 shows XRD patterns of the porous PTFE membranes before and after ALD deposition of alumina. The pristine membrane presents an intense peak centered at 17.81 corresponding to the diffraction from (1 0 0) planes with a spacing of 4.9 A˚ in semicrystalline PTFE composed of lamellar crystals [25]. After 200 and 300 ALD cycles, no new peak was observed in the whole scanning range (101o 2y o 501) because ALD-deposited alumina films are generally amorphous in nature [26–29]. However, the PTFE diffraction peak displays a detectable left-shift after deposition, and the shift became more significant with increasing ALD cycles. This finding indicates that the unit cell parameters or interlayer space of PTFE was slightly enlarged subsequent to the ALD deposition. Recently, it has been revealed that the

Fig. 3. TG curves of PTFE membranes subjected to different alumina ALD cycles at 150 1C showing the weight ratio of deposited alumina: (a) 0 cycle, the pristine membrane, (b) 50 cycles, (c) 200 cycles, and (d) 500 cycles.

439

high reactivity of the ALD precursors can lead to an incorporation of transition metals into PTFE [19]. In the case of ZnO ALD using Zn(C2H5)2 (DEZ) as the Zn precursor, remarkable structural changes and even phase transformation of PTFE were found after the incorporation of Zn into PTFE, in which the point defects and chain-end parts composed of –CF3 were proposed as the dominating reactive sites for Zn infiltration [19]. It is reasonable to presume that during the ALD deposition of alumina on our PTFE membranes, a similar intercalation of Al into the PTFE lattice or interlayer also happened. Therefore, the initial ALD deposition of alumina on PTFE cannot be simply described as formation of a physically bonded interface.

3.4. Adhesion strength of the deposited alumina on the PTFE substrate According to the initial ALD deposition mode of the alumina layer on PTFE membranes, the adhesion of the alumina film on PTFE might be poor due to the non-bonded, insufficient and shallow anchoring sites in the polymer. This is the reason why the ALD oxide films on smooth PTFE sheets failed to pass the Scotch tape test, as performed by Kemell et al. [18]. We evaluated the adhesion force between the deposition layer and the PTFE by measuring the percentage of detached alumina subjected to an aggressive ultrasonication treatment. It was found that the membrane treated with 50 ALD cycles showed a 2.7% loss of the total amount of deposited alumina after ultrasonication at a power of 300 W for 10 min. For the membrane subjected to 500 cycles, only less than 0.05% of the deposited alumina fell off from the membrane after the same ultrasonication challenging. SEM examinations showed there was no noticeable morphology change for the membrane subjected to 500 ALD cycles before and after the ultrasonication (Fig. 5), confirming the deposited alumina survived from the harsh ultrasonication treatment. However, nonporous PTFE films subjected to 50 and 500 alumina ALD cycles prepared under identical deposition conditions lost 41.5% and 6.2% of their deposited alumina, respectively. Therefore, in contrast to the poor adhesion between deposited alumina and nonporous PTFE films, alumina layers exhibited good adhesion to porous PTFE membranes. We attribute the good adhesion to the nanofibril-containing, porous structure of the PTFE membrane. In the presence of a large amount of PTFE fibrils bridging a three-dimensional porous configuration, ALD takes place throughout the entire thickness of the membrane, and alumina nucleates in the near-surface region and grows on the surface of the PTFE nanofibirls, forming a correspondingly 3D connected alumina layer wrapping PTFE nanofibrils. The 3D structure interwoven with the PTFE substrate can provide an additional spatial stabilizing force for the deposited layer. Revisiting this alumina ALD process on the PTFE microfiltration membranes, there are two features worth special concerns. First, PTFE membranes have a netlike porous structure. Second, an exposure process was always applied between the dosing and the purging process. It is known that a sufficient TMA uptake is associated with a large diffusion rate for TMA into the polymer film [17]. The highly porous PTFE membrane is supposed to provide more nucleation sites for TMA due to its large effective surface area. In addition, the exposure process not only promotes the sufficient diffusion of TMA into the hierarchical pores, it also supplies enough time for the TMA to penetrate the PTFE surface and encounter free volumes to achieve retention. Consequently, more TMA molecules in the deeper subsurface regions can survive the following purging process, finally producing more alumina tightly embedded in the PTFE. Furthermore, a long exposure time can also facilitate the possible chemical coordination of Al on the reactive sites of PTFE for nucleation, as we discussed above. Accordingly, the

Fig. 4. XRD patterns of PTFE membranes subjected to different alumina ALD cycles at 150 1C: (a) native, (b) 200 cycles, and (c) 300 cycles.

440

Q. Xu et al. / Journal of Membrane Science 415–416 (2012) 435–443 characteristic IR peaks arising from alumina is proportional to the thickness of the deposited alumina layer. However, the alumina IR peaks are very weak and it is technically difficult to compare the peak intensity from samples with changing thickness of deposited alumina layers. Alternatively, because of the shielding effect of alumina layer to PTFE substrates we can monitor intensity changes of PTFE characteristic peaks which are adequately strong to reflect the thickness of the alumina layer. Apparently, lower PTFE peaks indicate thicker alumina layers. Taking the membranes prepared at 150 1C with 50 and 100 cycles, respectively, as examples, we noticed that the peak intensity of 100-cycle-coated membrane was much weaker than the one subjected to 50 cycles because of the much thicker alumina layer deposited on the former. As shown in Fig. 7, in the wave number range of 500–800 cm  1, there are a number of strong peaks in the pristine PTFE membrane and these peaks remain but become weaker after alumina deposition under different deposition conditions. We compare the intensity of the strongest peak at 555 cm  1 attributed to the bending vibrations of CF2 groups [33] from samples prepared at different temperatures. The intensity diminishes in the order of samples prepared at 250, 200, 80, and 150 1C, suggesting the thickness of the deposited alumina increases in the same order. As there is also an ordering of mass uptake of these samples obtained by TG analysis, we compared orderings obtained from two different approaches and found that they are in accurate consistency. 3.6. Hydrophilicity improvement of the alumina-coated PTFE membranes

Fig. 5. SEM images of the PTFE membrane subjected to 500 alumina ALD cycles before (a) and after (b) ultrasonication.

We then investigate the effect of hydrophilicity enhancement of alumina ALD on PTFE membranes. By measuring the static water contact angle of the pristine PTFE membrane, an initial contact angle of 1317 31 was obtained (Fig. 8). This hydrophobic behavior is mostly due to the internal chemical structure of PTFE. Moreover, the surface roughness of the porous structure composed of nanofibrils can make the contact angle further higher than that of flat PTFE surfaces [34]. It is known that alumina films are highly hydrophilic because of its higher surface energy and abundant surface hydrophilic groups. The contact angle slightly decreased to 126 7 31 for PTFE membranes subjected to 100 ALD cycles. This result indicated that the hydrophobic PTFE surfaces were not fully covered by alumina during this stage. After 200 ALD cycles, the contact angle decreased dramatically to 627 41, due to the formation of a continuous alumina film. Increasing the ALD cycles to 300 and 500 led to further improvement of the hydrophilicity of the porous PTFE membrane. Especially, for the membrane subjected to 500 cycles, the water droplet easily penetrated into the PTFE bulk in the contact angle measurement, and disappeared within 10 s. Importantly, this hydrophicility improvement of the alumina-deposited PTFE membranes in stable for at least three months under ambient conditions since their values of water contact angle do not change compared to the newly deposited membranes. 3.7. Permeation and retention properties of the deposited PTFE membranes

alumina/PTFE adhesion was greatly enhanced in our case, which had not been achieved before.

3.5. The temperature effect of alumina ALD on PTFE membranes Deposition temperature plays a significant role in the ALD process since temperature influences the adsorption, diffusion in polymer matrix, as well as the reactivity of precursors on one hand, and changes the segmental mobility and flexibility of polymer chains which in turn alter the adsorption and diffusion behaviors of precursors inside the polymer. We carried out alumina ALD on PTFE membranes with fixed 100 cycles at 80, 150, 200, and 250 1C, respectively, to investigate the temperature effects. As shown in Fig. 6, alumina particulates appeared on the surface of all the samples prepared at different temperatures whereas the one prepared at 150 1C exhibited alumina particulates with biggest size. In addition, thermogravimetric analysis indicated that the amount of deposited alumina was 12.4%, 14.8%, 11.5%, and 10.2% for these samples, respectively, confirming the largest amount of alumina was deposited on the PTFE membranes deposited at the temperature of 150 1C. Higher temperatures enhance the reactivity and diffusion in the PTFE matrix of precursors while weaken the physical sorption of precursors. In the relatively low temperature region, enhancement of precursor reactivity and diffusion plays the dominating role, therefore, a higher amount of alumina was obtained at 150 1C than at 80 1C. Similarly, Jur et al. observed a larger mass uptake over 25 alumina ALD cycles on polypropylene films at 90 1C compared to 60 1C [30]. However, when the temperature increased further, poor adsorption of precursors and escape of AlOHn and AlCH3n surface species take effect. For example, it was observed that every growth cycle of alumina ALD decreases progressively with temperature between 177 1C and 300 1C [31,32]. As a consequence, lower amounts of deposited alumina were obtained when the depositing temperature increased from 150 1C to 200 1C and 250 1C. Therefore, as a result of these competing factors, alumina ALD takes place with the fastest rate at a moderate deposition temperature, e.g., 150 1C. FTIR measurements in the transmission mode were also performed to compare the thickness of the deposited alumina layer on PTFE membranes prepared at different deposition temperatures and cycles. The intensity of the

Fig. 9 presents a plot of pure water flux (PWF) of PTFE membranes as a function of ALD cycles. The PWF of the membrane subjected to 20 ALD cycles, increased dramatically and attained 67.7% higher than the pristine membrane. Likewise, the PWF of the membranes modified by 20, 50 and 100 cycles also exhibited a remarkable increase (a maximal increase of 69.2% of 100 cycles). As revealed in Fig. 5, although the hydrophilicity of the porous membranes just changed slightly during these ALD processes, the high efficiency of ALD in the upgrading of filtration performances was achieved. However, the PWF of the modified membranes began to decrease with the increasing number of ALD cycles after 100 cycles. For the membrane with 500 ALD cycles, a much lower flux with a 30.8% decline was obtained. We considered that the PWF of the membranes was synergistically influenced by the hydrophilicity of the pore surface as well as the pore size. When ALD deposition was performed with limited cycles (less than 100 cycles in our case), the increased hydrophilicity could overcome the negative effect due to the pore reduction by alumina deposition. With further increase in the number of ALD cycles, the pore reducing effect induced by the overgrowth of thicker alumina layers on the pore surface gradually dominated the water permeability through the membrane regardless of the further improved hydrophilicity, and hence decreased the water flux from its peak value. Therefore, the optimal ALD window for significantly increasing the PWF of the PTFE membranes is 20–100 cycles following the current ALD parameters. In contrast, the separation performance of the PTFE membranes was predominantly determined by the pore size. As shown in Fig. 9, the pristine membrane (average pore diameter of 200 nm) had an initial retention of 86.8% to polystyrene nanospheres with a size of 190 nm. After alumina deposition with 50 ALD cycles, the retention increased to 96.7% as a result of the reduced pore size. The retention performance almost kept unchanged with further increasing ALD cycles. 3.8. Fouling resistance of the deposited PTFE membranes A dynamic BSA filtration experiment was performed to evaluate the effect of hydrophilic modification on fouling resistance of PTFE membranes. Because the dynamic diameter of BSA ( 7.5 nm) [35] is much smaller than the mean pore

Q. Xu et al. / Journal of Membrane Science 415–416 (2012) 435–443

441

Fig. 6. Surface SEM images of PTFE membranes subjected to 100 cycles of alumina ALD at different temperatures: (a) 80 1C, (b) 150 1C, (c) 200 1C, and (d) 250 1C.

Fig. 7. FTIR spectra of the pristine PTFE membrane and PTFE membranes subjected to ALD alumina under different deposition conditions.

Fig. 8. Average water contact angles measured on PTFE membranes deposited with different ALD alumina cycles. The photographs of the water droplet on these membranes were also presented in the figure.

Fig. 9. Pure water flux and retention to monodispersed 190 nm polystyrene nanospheres of PTFE membranes subjected to alumina deposition of different ALD cycles.

Fig. 10. The relative flux ratio, F/F0, change of the pristine and modified membrane with filtration time: (a) the pristine membrane, (b) 100 cycles, (c) 300 cycles, and (d) 500 cycles. F0 and F stand for the flux in the filtration of pure water and the BSA solution, respectively.

442

Q. Xu et al. / Journal of Membrane Science 415–416 (2012) 435–443

diameter (  200 nm) of the pristine membrane, it is expected that the membrane have negligible retention to BSA based on size discrimination mechanism. It was found that the unmodified pristine PTFE membrane exhibited a dramatic decline in the relative flux ratio, F/F0, with filtration time (Fig. 10a). For example, F/F0 significantly reduced to 32.6% after 90 min’ running, suggesting serious membrane fouling occurs because of the preferential adsorption of BSA molecules to PTFE. The strong adhesion force between PTFE and BSA should be attributed to the hydrophobic interaction between PTFE and hydrophobic groups in BSA [36]. In contrast, alumina-deposited membranes displayed less profound flux reduction due to the presence of hydrophilic alumina on the surface (Fig. 10b–d). Furthermore, membranes subjected to more cycles of alumina deposition exhibit less reduction in flux. Values of F/F0 of membranes with 100, 300, and 500 ALD cycles are 36.3%, 52.6%, and 70.7%, respectively, after 90 min’ BSA filtration. The improvement in fouling resistance of membranes with different alumina cycles is in good consistence with the enhancement of hydrophilicity illustrated in Fig. 9, and we should attribute the improved fouling resistance mainly to the enhanced hydrophilicity of the deposited alumina layer. The suppression of the adsorption and accumulation of proteins on the modified membrane surface was mainly due to the unique coordination and hydration of the hydroxyl groups formed on the surface of alumina layer with surrounding water molecules. In addition, for the pristine and 100-cycle-deposited PTFE membranes, they showed a continuous trend of flux reduction during the entire period of 90 min’ running. However, for the membranes subjected to 300 and 500 ALD cycles, their flux kept reducing in the first 50 min then leveled off for longer BSA filtration, indicating a saturated adsorption of BSA was reached earlier for membranes subjected to higher numbers of ALD cycles. Recalling the fact that alumina particles are bigger on membranes with more ALD cycles, we should also attribute the improving fouling resistance of thickly deposited PTFE membranes to their smoother surfaces which adsorb less protein in addition to their enhanced hydrophilicity.

4. Conclusions In this work we used ALD deposition of alumina as an example to demonstrate that ALD is an efficient tool to modify the surface properties and tune the pore size of porous PTFE membranes. We showed that the thickness of the deposited layer and the effective pore size could be well controlled by simply alternating the number of ALD cycles. An ultrasonication challenging evidenced that a strong binding force existed between the deposition layer and the PTFE substrate. Due to the employment of an exposure process for the ALD deposition, a mode combining subsurface nucleation with possible chemical coordination was proposed for the growth of alumina layer on the porous PTFE. Water contact angle measurement and pure water flux experiment indicated that the surface of PTFE could gradually evolve from hydrophobicity to hydrophilicity with increasing ALD cycles. Within an optimized ALD operation window, a significantly increased PWF of the PTFE membranes was achieved. This study provides a simple and mild route to modifying the highly inert and stable material PTFE. By exploiting diverse ALD processes, production of various metal oxide/PTFE hybrid structures with specific functions can be expected.

Acknowledgments This work is financially supported by the National Basic Research Program of China (2011CB612302), the National Natural Science Foundation of China (21176120), an open research grant of the State Key Laboratory of Materials-Oriented Chemical Engineering for (KL10-01) and the Project of Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD). References [1] H.A. Rigby, C.W. Bunn, A room-temperature transition in polytetrafluoroethylene, Nature 164 (1949) 583–583. [2] C.W. Bunn, E.R. Howells, Structures of molecules and crystals of fluorocarbons, Nature 174 (1954) 549–551.

[3] J.T. Dickinson, J.J. Shin, W. Jiang, M.G. Norton, Neutral and ion emissions accompanying pulsed excimer laser irradiation of polytetrafluoroethylene, J. Appl. Phys. 74 (7) (1993) 4729–4736. [4] J.G. Drobny, Fluoropolymers in automotive applications, Polym. Adv. Technol. 18 (2) (2007) 117–121. [5] W.E. Hanford, R.M. Joyce, Polytetrafluoroethylene, J. Am. Chem. Soc. 68 (1946) 2082–2085. [6] C. Girardeaux, Y. Idrissi, J.J. Pireaux, R. Caudano, Etching and functionalization of a fluorocarbon polymer by UV laser treatment, Appl. Surf. Sci. 96–98 (1996) 586–590. [7] M.A. Bruk, Radiation-thermal crosslinking of polytetrafluoroethylene, High Energy Chem. 40 (6) (2006) 357–369. [8] M.K. Sohail, R. Frankeb, U. Gohsa, Friction and wear behaviour of electron beam modified PTFE filled EPDM compounds, Wear 266 (1–2) (2009) 175–183. [9] M. Tutiya, Nuclear magnetic resonance absorption of polytetrafluoroethylene gmma-irradiated at high temperature, J. Appl. Phys. 11 (10) (1972) 1542–1546. [10] N.I. Bol’shakova, V.S. Tikhomirov, V.I. Serenkov, I.M. Abramova, High-temperature radiolysis of poly(tetrafluoroethylene), Vysokomolekulyarnye Soedineniya, Seriya B: Kratkie Soobshcheniya 17 (8) (1975) 572–574. [11] C.Y. Tu, Y.L. Liu, J.Y. Lai, Surface grafting polymerization and modification on poly(tetrafluoroethylene) films by means of ozone treatment, Polymer 46 (18) (2005) 6976–6985. [12] W.H. Yu, E.T. Kang, K.G. Neoh, Controlled grafting of comb copolymer brushes on poly(tetrafluoroethylene) films by surface-initiated living radical polymerizations, Langmuir 21 (1) (2005) 450–456. [13] N. Zettsu, H. Itoh, K. Yamamura, Surface functionalization of PTFE sheet through atmospheric pressure plasma liquid deposition approach, Surf. Coat. Technol. 202 (22–23) (2008) 5284–5288. [14] M.A. Mohammed, V. Rossbach, Surface activation of polytetrafluoroethylene by bonding of polymeric silicic acid, J. Appl. Polym. Sci. 50 (6) (1993) 929–939. [15] S.Y. Wu, E.T. Kang, K.G. Neoh, H.S. Han, K.L. Tan, Surface modification of poly(tetrafluoroethylene) films by graft copolymerization for adhesion improvement with evaporated copper, Macromolecules 32 (1) (1999) 186–193. [16] Y.L. Liu, C.C. Han, T.C. Wei, Y. Chang, Surface-Initiated Atom Transfer, Radical polymerization from porous poly(tetrafluoroethylene) membranes using the C–F groups as initiators,, J. Polym. Sci., Part A: Polym. Chem. 48 (10) (2010) 2076–2083. [17] C.A. Wilson, R.K. Grubbs, S.M. George, Nucleation and growth during Al2O3 atomic layer deposition on polymers, Chem. Mater. 17 (23) (2005) 5625–5634. ¨ ¨ Surface modification of thermo[18] M. Kemell, E. Farm, M. Ritala, M. Leskela, plastics by atomic layer deposition of Al2O3 and TiO2 thin films, Eur. Polym. J. 44 (2008) 3564–3570. [19] S.M. Lee, V. Ischenko, E. Pippel, A. Masic, O. Moutanabbir, P. Fratzl, M. Knez, An alternative route towards metal–polymer hybrid materials prepared by vapor-phase processing, Adv. Funct. Mater. 21 (16) (2011) 3047–3055. [20] S.M. Lee, E. Pippel, U. Goesele, C. Dresbach, Y. Qin, C.V. Chandran, T. Braeuniger, G. Hause, M. Knez, Greatly increased toughness of infiltrated spider silk, Science 324 (5926) (2009) 488–492. [21] Z.Y. Yu, L. Chen, Su. Chen, Uniform fluorescent photonic crystal supraballs generated from nanocrystal-loaded hydrogel microspheres, J. Mater. Chem. 20 (29) (2010) 6182–6188. [22] S.K. Karuturi, L.J. Liu, L.T. Su, Y. Zhao, H.J. Fan, X.C. Ge, S.L. He, A.T.I. Yoong, Kinetics of stop-flow atomic layer deposition for high aspect ratio template filling through photonic band gap measurements, J. Phys. Chem. C 114 (35) (2010) 14843–14848. [23] M.I. Dafinone, G. Feng, T. Brugarolas, K.E. Tettey, D. Lee, Mechancial reinforcement of nanoparticle thin films using atomic layer deposition, ACS Nano 5 (6) (2011) 5078–5087. [24] S. Lasserre, J. Derouault, Cryoscopic study and vibrational characterization of the polyalumoxans or alumina aggregates formed during the hydrolysis of trimethylaluminum in benzene solutions, New J. Chem. 7 (11) (1983) 659–665. [25] J. Guo, L.P. Wang, S.C. Wang, J. Liang, Q.J. Xue, F.Y. Yan, Preparation and performance of a novel multifunctional plasma electrolytic oxidation composite coating formed on magnesium alloy, J. Mater. Sci. 44 (8) (2009) 1998–2006. [26] M. Kang, J.S. Lee, S.K. Sim, B. Min, K. Cho, H. Kim, M.Y. Sung, S. Kim, S. Song, M.S. Lee, Structural and optical properties of as-synthesized, Ga2O3-coated, and Al2O3-coated GaN nanowires, Thin Solid Films 466 (1–2) (2004) 265–271. [27] K. Goodman, V. Protasenko, J. Verma, T. Kosel, G. Xing, D. Jena, Molecular beam epitaxial growth of gallium nitride nanowires on atomic layer deposited aluminum oxide, J. Cryst. Growth 334 (1) (2011) 113–117. [28] L.J. Antila, M.J. Heikkila, V. Makinen, N. Humalamaki, M. Laitinen, V. Linko, P. Jalkanen, J. Toppari, V. Aumanen, M. Kemell, P. Myllyperkio, K. Honkala, H. Hakkinen, M. Leskela, J.E.I. Korppi-Tommola, ALD grown aluminum oxide submonolayers in dye-sensitized solar cells: the effect on interfacial electron transfer and performance, J. Phys. Chem. C 115 (33) (2011) 16720–16729. [29] A. Philip, S. Thomas, K.R. Kumar, Explanation for the appearance of alumina nanoparticles in a cold wall atomic layer deposition system and their characterization, Vacuum 85 (3) (2010) 368–372.

Q. Xu et al. / Journal of Membrane Science 415–416 (2012) 435–443

[30] J.C. Spagnola, B. Gong, S.A. Arvidson, J.S. Jur, S.A. Khan, G.N. Parsons, Surface and sub-surface reactions during low temperature aluminium oxide atomic layer deposition on fiber-forming polymers, J. Mater. Chem. 20 (2010) 4213–4222. [31] A.W. Ott, J.W. Klaus, J.M. Johnson, S.M. George, Al2O3 thin film growth on Si(1 0 0) using binary reaction sequence chemistry, Thin Solid Films 292 (1– 2) (1997) 135–144. [32] A.C. Dillon, A.W. Ott, J.D. Way, S. George, Surface chemistry of Al2O3 deposition using Al(CH3)3 and H2O in a binary reaction sequence, Surf. Sci. 322 (1–3) (1995) 230–242.

443

[33] L.N. Ignat’eva, V.M. Buznik, IR-Spectroscopic examination of polytetrafluoroethylene and its modified forms, Russ. J. Gen. Chem. 79 (3) (2009) 677–685. [34] R. Blossey, Self-cleaning surfaces-virtual realities, Nat. Mater. 2 5 (2003) 301–306. [35] T.R. Gaborski, J.L. Snyder, C.C. Striemer, D.Z. Fang, M. Hoffman, P.M. Fauchet, J.L. McGrath, High-performance separation of nanoparticles with ultrathin porous nanocrystalline silicon membranes, ACS Nano 4 (11) (2010) 6973–6981. [36] M. Hashino, K. Hirami, T. Ishigami, Y. Ohmukai, T. Maruyama, N. Kubota, H. Matsuyama, Effect of kinds of membrane materials on membrane fouling with BSA, J. Membr. Sci. 384 (1–2) (2011) 157–165.